Wireless communication between mobile users is becoming more and more popular as devices and technology is developed. The infrastructure roll out is expanding within both telecom systems and data network systems. Today also the telecom systems are increasingly using packet switched networks and the trend is clear towards this scheme of packet based routing. This system has been used for many years in the data networks and thus many standardized routing protocols exist for this purpose. However, they are not prepared for rapidly changing network topographies like for instance so called ad hoc networks.
Wireless ad hoc networks are characterized in that they do not have the same static nature as an ordinary wired network infrastructure, the ad hoc based network do not have a centralized control and is often created in a spontaneous manner. It maintains control through a decentralized concept. Nodes can be connected or disconnected in an uncontrolled manner as compared to standard fixed network architectures; the nodes may come and go quickly which leads to a dynamically changing network topology. In some cases such ad hoc networks are formed by user/client devices themselves as infrastructure components. These components are then truly mobile in the sense that the users move around, in and out of a network cell, and therefore the infrastructure will move around and change dynamically accordingly. This is an exciting and promising way of building an infrastructure; however, it sets very high demands on the routing protocol.
Other problems in a wireless environment are due to radio specific questions that will degrade the performance and efficiency of the network flow. There may be fading problems due to the movement of infrastructure nodes or movement of objects in the radio environment, and there may be problems due to interference from other radio sources within range.
These kinds of network topographies have been used in the military environment but are now migrating into the civilian area as well. Wireless systems are now used to rapidly build infrastructure areas for, e.g. wireless broadband access in residential areas or commercial areas. It may be used for temporary infrastructure build up, for example in an emergency situation, in a disaster area, or on the battlefield for military purposes. It could also be used to build up temporary access coverage areas during events like, for example, concerts, conferences, meetings, or seasonal tourist areas. In these kinds of areas, it is not necessary to have coverage all year around but only during specific periods, thus a fixed infrastructure build up in such a case may prove to be too expensive.
Today, several Internet Service Providers (ISP) offer wireless access at public or semi-public areas such as airports, restaurants, coffee shops, and hotels using fixed wireless infrastructure systems. These systems are often referred to as so called hotspots.
As the demand from the users to gain access increases considering coverage and bandwidth, one way of expanding the area of wireless coverage or bandwidth is to install more infrastructure components, however doing this with normal fixed wireless components are expensive and thus the idea to build networks using wireless routers has emerged. In this case ad hoc routing protocols may be used to have a simplified installation procedure.
There are basically two kinds of network usages when discussing ad hoc networks; the first one is the build up of a local area network without any external gateway providing access to an external network, for example Internet. This scheme may be found in installations concerning disaster areas or military installations on the battlefield. The other and probably more common usage is when one or several gateways provide the network with external connections to, for example, an IP (Internet Protocol) based network, private or public, e.g. Internet. In such a network configuration, data packets may take different routes and/or use different gateways depending on, for example, the data traffic type, congestions, or routing cost.
Packet based routing schemes often build their communication network systems around a layered model, for instance the OSI reference model. The communication software or hardware is divided into several smaller sub units, layers, working in a hierarchical manner. Information and communication control parameters are passed up and down locally and between the same layers between the sending and receiving ends. Each such layer is responsible for different tasks in the communication order. In respect to routing the first three layers according to the OSI reference model are the most important.
Layer 1 is responsible for the physical transmission of bits of data; examples of physical means may be, for instance, the wired link in an Ethernet based network or a wireless link in a Wireless Local Area Network (WLAN).
Layer 2 is often called the Link layer or the MAC layer and is responsible of transmitting chunks of data, error detection, and network resource coordination.
Layer 3 is often called the Network layer; it is responsible for enabling communication between any pair of nodes in a network. This layer takes, for example, care of routing calculations and some congestion control. For this purpose different routing protocols has been developed depending on the type of network.
Packet routing protocols in the IP-based networks are generally based on routing algorithms using distance vector or link state information to find and maintain a route for each pair of source and destination nodes in the network. In principle, in the distance vector routing algorithms, each router broadcasts the distance to all hosts to its neighbor routers, and each router receiving the information calculates the shortest route to each of the hosts in the network. In the link-state routing algorithms, each router broadcasts the status information of each of its adjacent network links to its neighbor routers, and each router receiving the information maintains the database of the entire picture of the network from the link status information and calculates the shortest route to each host based on the link costs in the database. These routing algorithms are designed for relatively static networks and thus new routing algorithms must be designed for ad hoc networks whose topology changes frequently.
There are basically two categories of existing routing protocols for ad hoc networks. These are “proactive” (table driven) and “reactive” (on-demand) routing protocols. Protocols having combinations of these protocols are also possible.
Proactive routing protocols constantly and periodically calculate a route to all hosts in the ad hoc network, and thus a route is always available when a packet needs to be sent to a particular destination host. The results are kept in routing tables in all nodes.
In order to maintain routes to each host, control messages are exchanged among the routers to notify changes of the network configuration and link status. Distance vector and link state routing protocols are both categorized as proactive protocols. It should be noted that control messages lead to overhead and may result in reduced network efficiency. Also, the proactive protocols may have difficulty in maintaining valid routes when the network topology changes frequently.
DSDV (Destination-Sequenced Distance Vector Routing) is a proactive routing protocol based on the distance vector algorithm, adapting the Routing Information Protocol (RIP) to ad hoc networks. Each node maintains a routing table in which the node stores the next hop node and hop count to each of all the reachable destination hosts. In DSDV, each node broadcasts or multicasts routing updates periodically, or when it detects changes of the network topology. Incremental updates, which update only information about changes since the last update, are also used in order to reduce control traffic.
A reactive protocol only performs control message exchange to find/update a route when there is a data packet to be sent. When a source node wants to send data packets, it initiates the control protocol to find a route by sending a route request message to its neighbor nodes. By this principle, the reactive approach is good in that network resources are not wasted when there are no packets to be transported. However, it takes longer time to send packets when a route has to be formed for the first time. AODV and DSR are representative reactive protocols.
AODV (Ad hoc On-Demand Distance Vector Routing) protocol uses the DSDV algorithm and creates/updates routes on an on-demand basis, that is, only when a source node wants to send a data packet. This leads to reduction of the number of required broadcasts for finding/updating a route.
In AODV, each node maintains a list of detected neighbor nodes. The neighbor list is updated in one of the following three ways: a) when a packet is received from the neighbor node, b) by receiving local advertisement, that is, hello message, from the neighbor node, or c) through feedback from the link layer. Hello messages are broadcasted periodically from each node to its neighboring nodes to inform them about its presence.
In AODV, each node maintains a routing table for all the destinations, each of which the node is either communicating with or forwarding data packets to on behalf of other nodes. For each destination, there is an entry in the routing table that contains information about the destination, such as the IP address, the sequence number for the destination node, hop count to the destination, the next hop node to the destination, and lifetime for the route.
When a node wants to communicate with a destination node, that is, to send data packets to the destination, then the source node initiates a route discovery mechanism, where the source node broadcasts a route request (RREQ) to all detected neighbor nodes. When the neighbor node receives the RREQ message and has the entry for a fresh enough route to that destination in its routing table, then it sends back a route reply (RREP) message to the source node. If the neighbor node does not find a route entry for that destination, then it forwards the RREQ message to its own detected neighbor nodes. When the destination node receives the RREQ, it returns the RREP message to the source node.
In the process of forwarding the RREQ packet, each intermediate node records the IP address of the neighbor node from which the first copy of the broadcast RREQ is received, by which a reverse route is established. The copies of the same RREQ messages received later are all discarded. The intermediate nodes add an entry to their routing table for the destination, where the neighbor node from which the RREP was received is recorded as the next hop node for that destination. The destination sequence number and lifetime of the route are copied from the RREP and recorded in the entry. When the RREP message is returned to the source node finally, a forward route from the source to destination is formed.
When a node detects that a route becomes unavailable by failure of the incident link on the route, it sends a route error (RERR) message to all the neighbor nodes, which use the route. The RERR message is sent on to their neighbor nodes and so on until it reaches the source node. The source node can then decide to either stop sending data packets or initiate a new route discovery.
DSR (Dynamic Source Routing) protocol uses a source routing mechanism in which the source node determines the complete sequence of nodes along the route on an on-demand basis and sets the list of the intermediate nodes in the packet header to indicate the sequence of nodes for the route. In this way, each packet has to carry the overhead for packet routing. However, the intermediate nodes do not need to maintain any information about the route and they can learn routes when delivering data packets.
In DSR, each node stores (caches) the routes it has learned. When a source node wants to send data packets to a destination node and has no entry in the cache for that destination, then it initiates a route discovery mechanism by broadcasting a RREQ message on its link-layer. Each node receiving the RREQ message appends their IP addresses to the RREQ message and then forwards it further. This process is done until the route to the destination is found or another node can provide a route to the destination node. Then a route reply (RREP) message containing the sequence of network hops to the destination node is returned to the source node.
In DSR, when a link failure is detected at a node (i.e. when the packet has been retransmitted a maximum number of times), that node removes the link from its routes cache and sends a route error (RERR) message to each of the nodes that have used that link since an acknowledgement was last received. Those nodes have to remove the routes including that link. The retransmission of the data packet from the source node is then handled by upper layers such as the Transmission Control Protocol (TCP).
One problem with the TCP/IP protocol is that it only makes use of IP addresses and in a data link in an Ethernet or token ring example, the network components has its own addressing scheme to which any network layer using the data link must conform. In, for example, an Ethernet several different network layers can cooperate at the same time, several network applications can use the same physical cable. When an Ethernet frame or packet is sent from one location to another it uses a 48 bit Ethernet address for determining the destination and source of the packet. A unique 48 bit Ethernet address is found in all Ethernet networking hardware and often called the MAC address (Media Access Control). This 48 bit address can be compared with the 32 bit IP address used in IPv4 (Internet Protocol version 4). Address resolution provides the mapping scheme between the two different forms of addresses. This mapping is done by the ARP (Address Resolution Protocol). ARP provides the mechanism to dynamically map hardware MAC addresses to IP addresses in a temporary memory space called the ARP cache. In other words the ARP translates the IP address to a MAC address.
The basic operation of ARP is as follows: when the IP layer wants to communicate with another device on the network, it checks the ARP cache (to see if there is a match with an Ethernet address). If there is no matching entry in the ARP cache, an ARP broadcast datagram is sent out that basically says: “The device with this IP address, reply with your Ethernet address?”. The receiving station (that has the IP address) responds with an ARP datagram that says: “This is my IP address and here is my Ethernet address”, The ARP cache is updated and the original IP layer information is passed on to the MAC layer for processing.
This scheme work fine for fixed line networks, but in wireless ad hoc and/or multihop networks not all network units can hear each other, in this case the standard ARP solution will not suffice. For this purpose some different layer 2.5 solutions for mapping IP and MAC addresses have been developed.
There are mainly two different routing protocols working as intermediary between layer 2 and 3, so called layer 2.5 protocols:                1. Lightweight Underlay Network (LUNAR). Instead of utilizing pure ARP, the ARP traffic is trapped by the Translating Underlay Network (SelNet) and rewritten to the extensible Resolution Protocol (XRP). XRP permits a much richer set of expressions than standard ARP. To maintain routes in LUNAR, all routes are cleared every third second and the path discovery procedure is redone.        2. Layer two DSR (the original version. In the route discovery process the packets used for routing is similar to ARP, but all addresses of the nodes along are recorded. As the protocol is a source routing protocol the whole route will also be added to every header of the data packets. The number of such addresses is limited and thereby results in a limited number of hops. The route maintenance is operating by detecting acknowledgements (ack.) hop by hop for each data packet transmitted. When no acknowledgement is achieved, the packet is retransmitted a certain number of times until the link is considered broken and another route is searched for.        
The problems with the above mentioned solutions can be summarized as follows:                1. Standard ARP can only be used for local peer-to-peer communication in a wireless network.        2. Packets in the network are normally not reused in ad hoc routing protocols.        3. LUNAR protocol is limited to a maximum of 10-15 nodes in total in a network cell and a maximum of 3 hops. In order to keep updated mapping tables it floods the network often, which results in additional control traffic. Also, it utilizes an external additional protocol to enable multihop functionality.        4. DSR on layer 2 is limited to a certain number of hops since all network addresses are stored in all packets (both data and control traffic packets). This solution demands additional resources and a new ARP like protocol is introduced.        
Accordingly the above mentioned solutions will not be transparent in a standard IP based network and will take up extra resources both from the available network resources and from the computing power of the involved routing components.